Estimating the grades of polluted industrial sites: use of categorical information and comparison with threshold values

N. Jeannée (1), Ch. de Fouquet (2)

(1) Centre National de Recherche sur les Sites et Sols Pollués, 930 Bd Lahure, BP 537, 59505 DOUAI Cedex, France, currently at Geovariances, 49bis av Franklin Roosevelt, 77212 AVON Cedex, France.
(2) Centre de Géostatistique, Ecole des Mines de Paris, 35 rue Saint-Honoré, 77300 FONTAINEBLEAU Cedex, France.

Abstract

The sampling of polluted sites often leads to inaccurate estimates, particularly because of the restricted number of samples, the importance of the sampling errors and the high spatial variability at small distance. Auxiliary information like the history of the site or qualitative measurements are of interest to improve the quality of the grade estimates. On a former coking plant, polluted by PAH (Polycyclic Aromatic Hydrocarbons) the relations between grades and soils information (presence of coal tar, smell, clay...) are examined. A sensitivity analysis shows the utility of this auxiliary information known at additional points compared to the univariate kriging of the grades.

The delineation of the zones to remediate is frequently carried out by selecting the areas where the estimated grades exceed the chosen remediation grade. If the estimation is subject to large uncertainties, this selection may generate bias. The estimation of the probability that the true grade is greater than the remediation value makes it possible to take into account the uncertainties associated to the estimated grades. Moreover, it is necessary to specify the support to be retained for this selection, which differs generally from the support of the samples. Neglecting this support effect leads to bias in the calculation of the soil volumes to remediate, as the proportion of the values exceeding a given grade varies with the support size (samples, blocks of various sizes). In this paper, conditional expectation and disjunctive kriging are compared for the estimation of the probability to exceed a threshold on blocks. The evaluation of polluted sites is then improved using a consistent methodology.